OpenAI's GPT-6 Sol and Luna cost half or less of GPT-5.6 prices

OpenAI released two cheaper GPT-6 models on September 22, 2026. According to TechCrunch, “Anthropic released a new version of Opus 5.5 just 90 minutes before OpenAI’s release.”

The two models

GPT-6 Sol is aimed at complex coding and agent work. Through the API it costs $2 per million input tokens and $10 per million output tokens.

GPT-6 Luna is built for fast, high-volume jobs. In OpenAI’s description, quoted by TechCrunch, it suits “high-volume tasks with a clear goal, like summarizing documents, extracting information, or answering quick questions.” It costs $0.10 per million input tokens and $0.50 per million output tokens.

TechCrunch reports that the new models cost half as much as the GPT-5.6 versions of Sol and Luna. An OpenAI spokesperson told VentureBeat the new rates are permanent prices, not a promotional or introductory offer.

OpenAI also makes a quality claim. On its internal factuality evaluation, the company says, GPT-6 Sol “makes about half as many mistakes as its predecessor.” OpenAI notes the evaluation uses conversations selected because users had flagged errors, which it says are not representative of typical usage.

The price gap in one line

Put the two launches side by side and the difference is plain. Anthropic’s new Opus costs $4 per million input tokens. OpenAI’s new mid-tier model costs $2, and its small model costs a twentieth of that.

The comparison is not like for like. Sol’s $2 and $10 pricing matches Anthropic’s mid-tier Claude Sonnet 5, VentureBeat notes, while Opus is Anthropic’s most capable and expensive tier, and the benchmark claims on each side come from the companies that sell the models. Still, for a developer who pays per token, the headline prices are the numbers that arrive on the invoice.

What the timing says

VentureBeat described Opus 5.5 as arriving hours before OpenAI’s scheduled announcement, and TechCrunch put the gap at 90 minutes. Either way, both launches landed on the same day, and the top AI labs are now competing on price as openly as on capability.

For developers and businesses paying per token, the practical result is that the same automated work got cheaper twice in one day. For the labs, it sets a floor that is hard to raise later: OpenAI has said on record that these prices are permanent.